OpenSeed – Autonomous AI creatures that find their own purpose

Agents are treated as continuous processes that sleep to consolidate observations into priority-tagged markdown (no vector DB) and inject hard rules back into prompts. Every nth sleep can trigger a self-evaluation that actually rewrites source code, leaving a git log as the creature's autobiography — it's a provocative, technically imaginative take on agent memory and evolution. That said, self-modifying code and long-lived agents raise reproducibility and safety questions the project will need to confront to scale beyond hobbyist experimentation.